Guidelines for Teaching C++

Version v0.1.0

SG20

1 Obtaining This Document: The Most Recent Version and Alternate Formats

The most recent version of this document is available as an online HTML document at: https://mdadams.github.io/sg20_guidelines_for_teaching_cpp/latest/.

The version of the document that you are currently reading is available in the following formats:

  1. online (HTML) format as a single large HTML document: https://mdadams.github.io/sg20_guidelines_for_teaching_cpp/v0.1.0/

  2. EPUB format: https://mdadams.github.io/sg20_guidelines_for_teaching_cpp/v0.1.0/guidelines.epub

  3. online (HTML) format, split across multiple HTML documents: https://mdadams.github.io/sg20_guidelines_for_teaching_cpp/v0.1.0/split/ [Note: The support for this format needs more work to look nice.]

Older versions of this document are also available. In general version ver is available at https://mdadams.github.io/sg20_guidelines_for_teaching_cpp/ver. For example, version v0.1.0 (assuming that this version exists) would be available at https://mdadams.github.io/sg20_guidelines_for_teaching_cpp/v0.1.0.

2 Disclaimer

This document is intended as a proof of concept to solicit feedback from others. This document is incomplete. This document likely has at least a few errors.

3 Context and Aim of This Guide

This document is intended to serve as a resource for instructors to assist in the preparation of courses on C++ in a variety of settings, including university, college, and industry environments. The main objectives of this document are as follows:

  • to provide guidelines for content to be covered by courses of various difficulty levels on C++ (e.g., topics and learning outcomes)
  • to note some common misunderstandings and problematic points that may be encountered when teaching particular topics
  • to suggest resources useful for teaching C++
  • to present examples of curriculum for specific courses

This document does not itself provide a curriculum for a single specific course, but is rather a set of guidelines that can be used to prepare curricula for a wide variety of courses that differ in focus and level of sophistication. (This said, however, some links to other documents with examples of curricula for specific courses may be included herein.)

4 Use of This Document

[NOTE: This document follows the same license model as the C++ Core Guidelines. The LICENSE document is taken verbatim from the C++ Core Guidelines.] This document is made available under a MIT-style license. In simple terms, this license permits copying, use, modification, and creation of derivative works. A copy of the license is included in the section LICENSE.

5 Contributing to This Document

Contributions to this document are welcome. If you would like to help with this project as a contributor, please read the section How to Contribute.

6 Organization of This Document

The various concepts (i.e., ideas) to potentially be covered are partitioned into modules/areas. A module/area is very broad in scope and consists of numerous topics. [Note: Can I suggest that we use the term “area” or “unit” instead of “module”? I think that these other choices are better and also avoid any potential confusion over what is meant by “module” (i.e., C++ term versus plain English term).]

For each module/area, topics related to the module/area are identified. Then, for each topic, learning outcomes are specified. In order to address a wide variety of courses on C++, each topic is addressed at three proficiency levels. These proficiency levels allow each topic to be covered at more than one level of detail. This allows target audiences with different background and learning objectives to be accommodated. The three proficiency levels are as follows:

  • foundational. This level gives the learner the idea that a facility exists, what benefits it offers, and the basic ways of using it. [Note: Isn’t this just “novice”/“beginner”?]

  • main. This level shows mainstream uses and techniques. For abstraction and organizational mechanisms it also demonstrates how to build them. Note that a main description typically cannot be written concisely without relying on the foundational sections for a variety of facilities. This is a major reason to separate foundational and main sections of the presentation of a facility. The main section should give the learner a basic (but not detailed) understanding of how a facility might be implemented so that the learner can have a first-order understanding of any costs involved. Avoid language lawyering in the main section. [Note: The term “main” is not very descriptive/helpful. Could I suggest using “intermediate”?]

  • advanced. This level gives information suitable for an expert. For most topics there is an expert level of knowledge that most programmers rarely need and techniques that require detailed understanding of language rules or library implementation. Delving into advanced topics early would be distracting and reinforce C++’s unfortunate reputation of being “expert only.” The advanced features often are “expert only” but the foundational and main descriptions should not be. Finally, we explicitly mention subjects that are considered too advanced for inclusion in the guidelines.

The remainder of this document is organized as follows. The various topics are listed grouped by module/area. In cases where a topic might be classified into more than one module/area, the topic is listed under the module/area of most direct relevance. This is done in order to avoid duplication of content. (In the case that a module/unit is equally relevant to multiple modules/areas, the decision of which to select is made by a proverbial coin toss.) The order in which modules/areas are presented is not meant to imply any order of coverage in a course. The order in which items are listed is essentially arbitrary.

7 Summary of Modules/Areas and Topics

In the sections that follow, the various modules/areas and topics are presented. There is one section per module/area. For each module/area, a table listing the various topics in that module/area is provided. The ID for a module/area is linked to the detailed coverage of that module/area that comes later in the document. In the table, a checkmark indicates that learning outcomes exist for the specific proficiency level. A dash indicates that no learning outcomes exist for the specific proficiency level (simply because there are not any, not because the information is missing). In the case that the information for a knowledge unit is completely missing, a “❌” symbol is used.

[NOTE: These topics are taken mostly from the SG20 GitHub repository. They are not intended to be complete in any sense. In fact, by gathering together all topics in one place where they are easily viewed, it is hoped that missing and unbalanced items will be more obvious.]

7.1 Compilation Model

ID Unit Foundational Main Advanced
[] Translation Units ? ? ?
[] Headers ? ? ?
[] Modules ? ? ?
[] Name Mangling ? ? ?
[] Phases of Translation ? ? ?
[] Separate Compilation ? ? ?
[] Linkage ? ? ?

7.2 Preprocessor

ID Unit Foundational Main Advanced
[] Preprocessor Metaprogramming ? ? ?
[] Inclusion ? ? ?
[] Macros ? ? ?

7.3 Basics Types, Objects, Values, Expressions, Statements, and Control-Flow Constructs

ID Unit Foundational Main Advanced
[] Constant Objects ? ? ?
[] Declarations and Definitions ? ? ?
[] Selection Constructs (e.g., if, ternary) ? ? ?
[] Looping Constructs (e.g., for, while, etc.) ? ? ?

7.4 Functions

ID Unit Foundational Main Advanced
[] Calling Functions ? ? ?
[] Parameter Passing (e.g., Passing By Value and Reference) ? ? ?
[func_args] Default Arguments ✔️
[] Returning Multiple Values ? ? ?
[] Overloading ? ? ?
[udl] User-Defined Literals ✔️ ✔️

7.5 User-Defined Types (Classes)

ID Unit Foundational Main Advanced
[] Special Member Functions ? ? ?
[copy_elision] Copy Elision and Implicit Moving (e.g., RVO and NRVO, etc.) ✔️ ✔️
[] Types ? ? ?
[] Conversions ? ? ?
[] Constructors and Destructors ? ? ?
[] Move/Copy Constructors and Assignment Operators ? ? ?
[] Member Functions ? ? ?
[] Sum Types ? ? ?
[] User-Defined Literals ? ? ?
[] Special Member Functions ? ? ?
[] Guidelines for Special Member Functions (e.g., Rule of Five, Rule of Zero) ? ? ?
[copy] Copy Semantics ✔️ ✔️
[] Moving and Copying ? ? ?
[] Lambdas ? ? ?

7.6 Inheritance and Class Hierarchies

ID Unit Foundational Main Advanced
[] Virtual Functions ? ? ?
[] Run-Time Type Information ? ? ?

7.7 Compile-Time Computation

ID Unit Foundational Main Advanced
[] Constant Expressions and Constant Evaluation ? ? ?
[] static_assert ? ? ?

7.8 Generic Programming (Templates)

ID Unit Foundational Main Advanced
[] Concepts ? ? ?
[] SFINAE ? ? ?
[] Template Metaprogramming ? ? ?
[] Function Templates ? ? ?
[] Requires Clauses ? ? ?
[req_expr] Requires Expressions ✔️ ✔️

7.9 Error Handling

ID Unit Foundational Main Advanced
[] Classes of Errors ? ? ?
[] errno ? ? ?
[] Error Codes ? ? ?
[] Exception Handling ? ? ?

7.10 Standard Library

ID Unit Foundational Main Advanced
[] Input/Output (I/O) ? ? ?
[] Containers, Iterators, and Algorithms ? ? ?

7.11 External (i.e., Non Standard) Libraries

ID Unit Foundational Main Advanced
[] Graphical User Interfaces ? ? ?

7.12 Building

ID Unit Foundational Main Advanced
[] Software Build Tools ? ? ?
[] Strategies for Handling Build Problems ? ? ?

7.13 Testing and Debugging

ID Unit Foundational Main Advanced
[] Source-Level Debuggers ? ? ?
[] Code Sanitizers ? ? ?
[] Test Frameworks ? ? ?
[] Debugging Strategies ? ? ?

7.14 Tooling

ID Unit Foundational Main Advanced
[] Compiler Toolchains ? ? ?
[] IDEs ? ? ?

7.15 Software Design

ID Unit Foundational Main Advanced
[] Design by Contract ? ? ?

8 Detailed Information for Modules/Areas and Topics

8.1 C++ object model: copy semantics

Skeleton instructions are typeset in italic text.

8.1.1 Overview

Provides a short natural language abstract of the module’s contents. Specifies the different levels of teaching.

Level Objectives
Foundational understanding how and when are copies made
Main implementing user-defined copy operations
Advanced special cases: copy elision

8.1.2 Motivation

Why is this important? Why do we want to learn/teach this topic?

Copy semantics allows the user to define how objects of a class get replicated and interact on a value level.

8.1.3 Topic introduction

Very brief introduction to the topic.

Explains when and how objects are copied.

8.1.4 Foundational: How and when are copies made

8.1.4.1 Background/Required Knowledge

A student is able to: * explain what a C++ type is? [C++ object model: types] * explain what an object is? [C++ object model: objects], [C++ object model: constant objects] * define and understand class invariants?

It helps when a student is able to: * use move semantics [C++ object model: move semantics] * explain special member functions [C++ object model: special member functions]

8.1.4.2 Student outcomes

A list of things “a student should be able to” after the curriculum. The next word should be an action word and testable in an exam. Max 5 items.

A student should be able to:

  1. explain what copy semantics accomplish
  • establishing “equivalent” object state in another object
  1. explain difference between copying a reference and copying a value*
  2. explain where copies are made

* In other languages these differences are sometimes referred to as shallow and deep copy.

8.1.4.3 Caveats

This section mentions subtle points to understand, like anything resulting in implementation-defined, unspecified, or undefined behavior.

  • Compiler-provided copy operations may result in ownership problems (e.g., char*). These ownership problems can generally be solved by using types whose copy operations have the appropriate semantics, e.g., std::string instead of char* to hold string values.

8.1.4.4 Points to cover

This section lists important details for each point.

  • Principle of copying
    • Copying of types, which follow the rule of zero
    • Copying of types, with user defined copy operations
    • Copying an object does not change the original
  • Practical applications
    • std::unique_ptr (has no copy)
    • Strings (copies the value)

8.1.5 Main: Implementing user-defined copy operations

8.1.5.1 Background/Required Knowledge

A student is able to: * identify special member functions [C++ object model: special member functions]

It helps when a student is able to: * use move semantics [C++ object model: move semantics] * explain the rule of zero [C++ object model: rule-of-zero] * explain the rule of five [C++ object model: rule-of-five]

8.1.5.2 Student outcomes

A list of things “a student should be able to” after the curriculum. The next word should be an action word and testable in an exam. Max 5 items.

A student should be able to: * explain when they have to implement the copy operations for their own type * Copy constructor * Copy assignment operator * implement copy operations for their own types * Optional: explain when copying with basic and strong exception guarantees is useful

8.1.5.3 Caveats

This section mentions subtle points to understand, like anything resulting in implementation-defined, unspecified, or undefined behavior.

  • Intricacies when implementing copy operations:
    • Examples of how not to write copy operations (e.g., C++03 std::auto_ptr)

8.1.5.4 Points to cover

This section lists important details for each point.

  • Copy constructors and copy assignment operators
    • How compiler generates default copy operations
    • =default, =delete (No copy)
    • How-to write your own copy operations
    • Rule-of-five
    • Copy assignment operators can be ref-qualified to avoid assigning into temporary objects.

8.1.6 Advanced

These are important topics that are not expected to be covered but provide guidance where one can continue to investigate this topic in more depth.

When can copies be elided and when does the standard guarantee copy elision. References: * Abseil tip of the Week #166 * cppreference - Copy elision

8.2 Functions: user-defined literals

Skeleton instructions are typeset in italic text.

8.2.1 Overview

Provides a short natural language abstract of the module’s contents. Specifies the different levels of teaching.

Level Objectives
Foundational using and understanding UDLs
Main implementing your own UDLs
Advanced Advanced use (“{}, {}!”_fmt(“Hello”, “World”))

8.2.2 Motivation

Why is this important? Why do we want to learn/teach this topic?

  • Allows clearer expression of intent in C++.
  • std::string: "Hello, world!"s
  • std::chrono: 3h + 10m + 5s

8.2.3 Topic introduction

Very brief introduction to the topic.

  • Explain the existence of user defined literals. Example: 12m + 17s is terse, expressive and type safe.

8.2.4 Foundational: Using UDLs

8.2.4.1 Background/Required Knowledge

A student: * knows how to form numeric literals, e.g., 1.5f means a float of value 1.5. * is familiar with the major C++ types: * bool (Boolean type) * int (Integer type) * double (Floating-point type) * std::string (Text type) * std::vector (Collection type) * knows that namespaces exist, and namespace std. * knows what using-declarations and using-directives are. [C++ object model: declarations]

8.2.4.2 Student outcomes

A list of things “a student should be able to” after the curriculum. The next word should be an action word and testable in an exam. Max 5 items.

A student should be able to:

  1. use using namespace std::string_literals[1].
  2. recognise UDLs in code that they are reading.
  3. figure out which UDL definitions for a used type exist.
  4. identify parts of the standard library that make use of UDLs.
  5. prevent the dangers of temporaries created with "blah"s as well as with std::string{"blah"}.
  6. effectively selects the right set of namespaces in using-directives from the sub-namespaces std::literals.

[1]: explain that it’s okay to use a using-directive to “activate” UDLs.

8.2.4.3 Caveats

This section mentions subtle points to understand, like anything resulting in implementation-defined, unspecified, or undefined behavior.

  • A student gets confused by the similarities and differences between built-in suffixes and UDLs and between UDLs from different namespaces.
  • A student “activates” two suffixes with the same signature from different namespaces.

8.2.4.4 Points to cover

This section lists important details for each point.

8.2.5 Main: implementing UDLs

8.2.5.1 Background/Required Knowledge

  • All of the above.

8.2.5.2 Student outcomes

A list of things “a student should be able to” after the curriculum. The next word should be an action word and testable in an exam. Max 5 items.

A student should be able to:

  1. write a UDL operator of their own.
  2. separate unrelated UDLs into distinct namespaces.

8.2.5.3 Caveats

This section mentions subtle points to understand, like anything resulting in implementation-defined, unspecified, or undefined behavior.

No caveats at present. #### Points to cover

This section lists important details for each point.

No caveats at present. ### Advanced {#udl-advanced}

These are important topics that are not expected to be covered but provide guidance where one can continue to investigate this topic in more depth.

8.3 Functions: default argument

Skeleton instructions are typeset in italic text.

8.3.1 Overview

Functions in C++ may be overloaded with different numbers and types of parameters. It may be of value to specify default arguments for some number of parameters, to allow a caller to avoid specifying arguments that rarely change, or to enable expanding the set of parameters while maintaining backward compatibility with existing callers.

Level Objectives
Foundational Define and use functions with default arguments
Main
Advanced refinement of default arguments through multiple declarations

8.3.2 Motivation

Default arguments allow the omission of arguments with obvious or common values. Also may be utilized to extend an existing function signature without forcing changes to existing calling code.

8.3.3 Topic introduction

Explain how default arguments work and how to define them.

8.3.4 Foundational: Using and defining functions with default arguments

8.3.4.1 Background/Required Knowledge

A student is able to:

8.3.4.2 Student outcomes

A student should be able to:

  1. Call to a function with a default argument with or without that argument specified
  2. Declare a function with a default argument, and omit the default in the definition’s signature
  3. Explain when the lifetime of a default argument begins and ends

8.3.4.3 Caveats

  • When no forward-declaration exists, the definition serves as the declaration
  • When multiple declarations exist, only one may specify the default for any particular parameter, but multiple declarations may specify the defaults for different parameters.
  • Additional default values may be specified for other parameters in repeat declarations
  • Calling an overloaded function with fewer arguments may be ambiguous with regard to an overload with default arguments

8.3.4.4 Points to cover

  • Default value may only be specified once for each parameter among all declarations
  • Default values must start from the rightmost parameter and continue leftward without gaps
  • Considerations of when to use default arguments vs overload set

8.3.5 Main: implementing *

8.3.5.1 Background/required knowledge

  • All of the above.

8.3.5.2 Student outcomes

A student should be able to:

8.3.5.3 Caveats

8.3.5.4 Points to cover

8.3.6 Advanced

Subsequent redeclarations of the same function may add default argument values, which are then usable by callers. Though a single parameter cannot be given a default argument twice in the same translation unit, it is legal, though ill-advised, to give the same function different default arguments in different translation units.

8.4 Module name: Requires Expressions

8.4.1 Overview

Level Objectives
Foundational Define and use requires-expressions to check satisfaction of expressions by given parameters
Main Define and use requires-expressions to check properties of expressions
Advanced

8.4.2 Motivation

Requires-expressions allow a developer to perform compile-time evaluation on the validity of other expressions. These are fundamental to the ability to write concepts. [Compile-time programming: concepts]

8.5 Topic introduction

Requires-expressions are compile-time predicates which evaluate to true when their specified set of expressions are all valid for a given set of inputs.

8.5.1 Foundational: Writing requires-expressions

8.5.1.1 Background/Required Knowledge

A student is able to:

It is helpful if:

  • The student is aware that attempting to specialize the template with types or values which do not match otherwise unstated assumptions will cause errors within the template.

8.5.1.2 Student outcomes

A student should be able to:

  1. Write a simple-requirement to assert the validity of an expression
  2. Write a type-requirement to check the existence of a type by its identifier
  3. Write a compound-requirement to test the resulting type of an expression
  4. Write a nested-requirement to test the constexpr value of an operation, as opposed to just the syntactic validity
  5. Use a requires-expression within a concept, requires-clause, or if constexpr condition

8.5.1.3 Caveats

To require that expressions, which evaluate to a boolean value like sizeof(t) == 4, evaluate to true a nested-requirement is needed (e.g., requires sizeof(t) == 4;). Omitting the requires results in a simple-requirement, which is satisfied based purely on syntactic validity, not on the result of the operation.

8.5.1.4 Points to cover

  • All requires-expression requirements terminate with a semicolon.
  • simple-requirements are used to check that an expression is well-formed.
  • nested-requirements are introduced with requires and primarily used to check the result of an expression computable by the compiler, including concepts or other requires-expressions.
  • type-requirements are introduced with typename and used to verify the existence of a type with a particular identifier.
  • compound-requirements are enclosed in braces and can be used to check the resulting type of an expression.
  • Checks are performed by the compiler, not at run time.
  • If covering usage of requires-expression with requires-clause, [Compile-time programming: requires clause] demonstrate requires requires and show how to ever avoid writing it by using a concept. [Compile-time programming: concepts]

8.5.2 Main: Advanced requirements

8.5.2.1 Background/required knowledge

  • All of the above.
  • Knowledge of noexcept

A student is able to:

8.5.2.2 Student outcomes

A student should be able to:

  1. Write compound-requirements which test the noexceptness of an expression.
  2. Use a concept as the target of a compound-requirement.

8.5.2.3 Caveats

8.5.2.4 Points to cover

  • Compound-requirements allow the optional ability to test whether an expression is marked as noexcept, by using a trailing noexcept keyword. ``` struct S { void foo() noexcept {} void bar() {} };

static_assert(requires(S s) { { s.foo() } noexcept; } ); // Succeeds. s.foo() is noexcept static_assert(requires(S s) { { s.bar() } noexcept; } ); // Fails. s.bar() is not noexcept ``* If the return-type-requirement of a compound-requirement is a concept, that concept is given the resulting type as the first parameter, followed by the specified parameters in the compound-requirement.{ ++x } -> Cwould substituteC<decltype((++x)), int>` and check that concept C is satisfied for those parameters.

8.5.3 Advanced

8.6 Copy Elision and Implicit Moving

8.6.1 Basic Learning Outcomes

8.6.2 Intermediate Learning Outcomes

After completing this unit, the student should be able to:

  • explain what copy elision and implicit moving are and why they are beneficial;

  • explain what is meant by return value optimization (RVO) and named RVO (NRVO);

  • for certain common code patterns, be able identify whether copy, move, or copy elision takes place.

8.6.3 Advanced Learning Outcomes

After completing this unit, the student should be able to:

  • identify the particular circumstances when copy elision is permitted (but not necessarily required) and when it is required;

  • in arbitrary code, be able to identify whether copy, move, or copy elision takes place.

8.6.4 Common Misunderstandings

Some common misunderstandings include the following:

  • Copy elision is only relevant to construction not assignment.

9 Examples of Course Curricula

[NOTE: Anyone have any suggestions of items to add here?] The following are examples of curricula for course on C++: …

10 License

[NOTE: This license is copied verbatim from the C++ Core Guidelines.]
Copyright (c) Standard C++ Foundation and its contributors

Standard C++ Foundation grants you a worldwide, nonexclusive, royalty-free,
perpetual license to copy, use, modify, and create derivative works from this
project for your personal or internal business use only. The above copyright
notice and this permission notice shall be included in all copies or
substantial portions of the project. This license does not grant permission
to use the trade names, trademarks, service marks, or product names of the
licensor, except as required for reasonable and customary use in describing
the origin of the project.

Standard C++ Foundation reserves the right to accept contributions to the
project at its discretion.

By contributing material to this project, you grant Standard C++ Foundation,
and those who receive the material directly or indirectly from Standard C++
Foundation, a perpetual, worldwide, non-exclusive, royalty-free, irrevocable,
transferrable license to reproduce, prepare derivative works of, publicly
display, publicly perform, and distribute your contributed material and such
derivative works, and to sublicense any or all of the foregoing rights to third
parties for commercial or non-commercial use.  You also grant Standard C++
Foundation, and those who receive the material directly or indirectly from
Standard C++ Foundation, a perpetual, worldwide, non-exclusive, royalty-free,
irrevocable license under your patent claims that directly read on your
contributed material to make, have made, use, offer to sell, sell and import
or otherwise dispose of the material. You warrant that your material is your
original work, or that you have the right to grant the above licenses.

THE PROJECT IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE PROJECT OR THE USE OR OTHER DEALINGS IN THE
PROJECT.

If you believe that anything in the project infringes your copyright, please
contact us at admin@isocpp.org with your contact information and a detailed
description of your intellectual property, including a specific URL where you
believe your intellectual property is being infringed.

11 Contributors

For a complete list of contributors, please refer to the repository containing this document on GitHub.

12 How To Contribute

Before attempting to contribute any content for consideration for inclusion in this document, please read the information below and and the referenced documents as appropriate.

All contributions to this project must be made in accordance with the license in section License. This document only offers guidance on teaching C++ as it is specified in the current version of the C++ standard. So, content should be presented relative to the most-recently ratified version of the standard. A detailed explanation of how to present the material for a topic is given in:

  • JC van Winkel, Bjarne Stroustrup, and Florian Sattler. P2193 — How to structure a teaching topic. https://wg21.link/p2193.

Any potential contributors should ensure that they read this document. The following document may also be helpful in providing some general background on the modular approach to teaching followed herein:

In order to prepare content, it is important to understand what learning outcomes are, and how to prepare good ones. Some information on learning outcomes can be found in the references. The following document offers a concise introduction to learning outcomes:

13 Glossary

Some possible terms to include in the glossary:

  • learning objective
    • learning outcome

14 References

14.1 References on Learning Outcomes

  • Christopher Di Bella, Simon Brand, and Michael Adams. P1389R0 — Standing Document for SG20: Guidelines for Teaching C++ to Beginners. https://wg21.link/p1389.

  • Christopher Di Bella. P1725R0 — Modular Topic Design. https://wg21.link/p1725.

  • JC van Winkel, Bjarne Stroustrup, and Florian Sattler. P2193 — How to structure a teaching topic. https://wg21.link/p2193.